Centroid-based summarization of multiple documents: sentence extraction, utility-based evaluation, and user studies

By Dragomir R. Radev, Hongyan Jing, M. Budzikowska. arXiv.org, 2000

589 citations recorded by the source corpus.

We present a multi-document summarizer, called MEAD, which generates summaries using cluster centroids produced by a topic detection and tracking system. We also describe two new techniques, based on sentence utility and subsumption, which we have applied to the evaluation of both single and multiple document summaries. Finally, we describe two user studies that test our models of multi-document summarization.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/ffb46f26ebff29b31b16c6e4b60d85019e19ac75

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

AgreeSum: Agreement-Oriented Multi-Document Summarization (2021, background): “…summarize the intersection of articles, a few past works have discussed the helpfulness of models that identify common infor- mation “centroids” among multiple related documents, so as to allow internet users to more efficiently understand events (Radev et al., 2000; Barzilay and McKeown, 2005).”
Introduction to the Special Issue on Summarization (2002, background): “…sentences (a lengthy solution), some approaches will select a representative passage to convey information in each cluster (Radev, Jing, and Budzikowska 2000), whereas other approaches use information fusion techniques to identify repetitive phrases from the clusters and combine the phrases into…”
Text Summarization Techniques: A Brief Survey (2017, background): “To accomplish this goal, two metrics are defined [25]: clusterbased relative utility (CBRU) and cross-sentence informational subsumption (CSIS).”
LexRank: Graph-based Lexical Centrality as Salience in Text Summarization (2004, methodology): “Several rerankers are implemented in MEAD, including one based on Maximal Marginal Relevance (MMR) (Carbonell & Goldstein, 1998) and the default reranker of the system based on Cross-Sentence Informational Subsumption (CSIS) (Radev, 2000).”
Opinosis: A Graph Based Approach to Abstractive Summarization of Highly Redundant Opinions (2010, methodology): “Also, since the position in text parameter is irrelevant in our case, we could easily turn this off with MEAD.”
Social Media Text Classification under Negative Covariate Shift (2015, methodology): “Alternative document representations have been proposed in the past and have been shown to perform well in many applications (Radev et al., 2000; He et al., 2004; Lebanon 2006; Ranzato and Szummer, 2008, Wang and Domeniconi, 2008).”
Inferring Strategies for Sentence Ordering in Multidocument News Summarization (2002, methodology): “This strategy was originally implemented in MultiGen and followed by other summarization systems (Radev, Jing, & Budzikowska, 2000; Lin & Hovy, 2001).”
Automatic Text Summarization: Past, Present and Future (2013, background): “In the case of generic summarization, computing similarity between sentences and the centroid of the documents to summarize has resulted in competitive summarization solutions [55, 63].”
Temporal summaries of new topics (2001, background): “Other work on news summarization, including work that uses the TDT corpora, focuses on single or multi-document summarization [25, 34, 12] of the stories, without attempting to capture the changes over time.”
A Survey on Automatic Text Summarization (2007, methodology): “In (Radev et al., 2000), two metrics are defined that resemble the two summands in the MMR (see Section 3.2): clusterbased relative utility (CBRU) and cross-sentence informational subsumption (CSIS).”